Post by c64 on Sept 11, 2017 16:09:45 GMT
I try to explain why professional keyboards are so expensive:
Inside the ear, there are tiny reeds responding to specific frequencies. Any waveform can be represented by a pattern of mixed sine waves of different frequency and intensity. In theory, a square wave is an infinitive number of sine waves mixed together. But since the number of different reeds in the ear is limited, there are not that many frequencies you need to mix to hear any possible sound pattern you can actually distinguish.
Simple electronic organs contain just a few sine wave generators you mix by pressing keys. With 8 oscillators per octave, any pressed single key sounds almost perfect, you can simulate almost any instrument in a good quality. When pressing two keys in the same octave, the result is still very acceptable. With 3 or more keys pressed, the result turns ugly fast. The keyboards for children from the 1980s and 1990s contain just 5 oscillators per octave, sometimes just 4 or even less!
A professional keyboard needs at least 10 oscillators per octave and the keyboard has many octaves. This is what makes them incredible expensive. All of them need to be in sync which requires a lot of circuity. The required space is immense!
You can see the massive space for the electronics in the pictures. And even with these bulky circuits, those organs are not very good by today's standards!
Yamaha came up with a single chip housing way over a hundred of oscillators all perfectly in sync and free of maintenance. This chip originally had cost thousands of dollars and is still very expensive. But it allows to press as many keys per octave as can make sense in music and you can have a lot of octaves (or use a second chip for even more). Organs using this chip lack the bulky boxes.
And here is where the Roland MT-32 had an impact on quality, size and price. Instead of containing hundreds of oscillators to compose the waveforms of tunes, it works mostly digital. It contains a - for back then standards - insane amount of RAM which serves as a ring buffer. The contents is constantly played by a D/A converter. When a MIDI command comes in, the proper waveform is taken out of a ROM, re-sampled by interpolation to match the desired frequency and then dumped into the ring buffer to be played. Back then when 64kB of RAM was considered as "decadent", this would be awfully expensive. Instead, there is a trick:
A musical tune can be split into 2 parts:
1. "Attack", the first part of the tune
2. "Sustain", the part where the tune is held and fading out.
The most important part to tell the difference between different sounds is the "Attack" phase, once the brain has identified what the tune sounds like, it is still happy with a less complex sustain phase.
So what the ROLAND MT-32 does is to keep tiny digital snippets of almost 100 instrument sounds in a ROM to create any complexity for the attack phase necessary to reproduce a very realistic sound. After the attack phase, the MT-32 then plays "plain" tunes out of simple digital oscillators for the Sustain phase.
This is nothing new, the Roland D-50 did that, too. But the MT32 can handle 5 independent instrument channels. Also back then the beginning of home computing made digital chips and especially RAM and ROM much more inexpensive so the MT-32 not only appeared to be as good as hundreds of expensive oscillators bundled into a big box (or highly expensive chip), it did even better (5 independent channels) and had only cost a fraction!
So from todays point of view, the MT32 is not very good. You can synthesize hundreds of digital channels with the full tune reproduced instead of just the attack phase with very inexpensive hardware nowadays. But since the MT-32 made the impossible possible back in 1982, it became a legend and so it is still traded with its weight in gold.
You can tell the limits of the MT-32 easily. E.g. a real piano sound requires a complex pattern over the entire length of the tune, the reason why the standard piano sound is very unrealistic played on the MT32. There are other instruments where the attack phase is just not enough to make it sound real. But with almost 100 instruments to choose from (and most very realistic), who cared back then? It was cheap and apparently you got much more than a Pro had for a fraction of the money.
Inside the ear, there are tiny reeds responding to specific frequencies. Any waveform can be represented by a pattern of mixed sine waves of different frequency and intensity. In theory, a square wave is an infinitive number of sine waves mixed together. But since the number of different reeds in the ear is limited, there are not that many frequencies you need to mix to hear any possible sound pattern you can actually distinguish.
Simple electronic organs contain just a few sine wave generators you mix by pressing keys. With 8 oscillators per octave, any pressed single key sounds almost perfect, you can simulate almost any instrument in a good quality. When pressing two keys in the same octave, the result is still very acceptable. With 3 or more keys pressed, the result turns ugly fast. The keyboards for children from the 1980s and 1990s contain just 5 oscillators per octave, sometimes just 4 or even less!
A professional keyboard needs at least 10 oscillators per octave and the keyboard has many octaves. This is what makes them incredible expensive. All of them need to be in sync which requires a lot of circuity. The required space is immense!
You can see the massive space for the electronics in the pictures. And even with these bulky circuits, those organs are not very good by today's standards!
Yamaha came up with a single chip housing way over a hundred of oscillators all perfectly in sync and free of maintenance. This chip originally had cost thousands of dollars and is still very expensive. But it allows to press as many keys per octave as can make sense in music and you can have a lot of octaves (or use a second chip for even more). Organs using this chip lack the bulky boxes.
And here is where the Roland MT-32 had an impact on quality, size and price. Instead of containing hundreds of oscillators to compose the waveforms of tunes, it works mostly digital. It contains a - for back then standards - insane amount of RAM which serves as a ring buffer. The contents is constantly played by a D/A converter. When a MIDI command comes in, the proper waveform is taken out of a ROM, re-sampled by interpolation to match the desired frequency and then dumped into the ring buffer to be played. Back then when 64kB of RAM was considered as "decadent", this would be awfully expensive. Instead, there is a trick:
A musical tune can be split into 2 parts:
1. "Attack", the first part of the tune
2. "Sustain", the part where the tune is held and fading out.
The most important part to tell the difference between different sounds is the "Attack" phase, once the brain has identified what the tune sounds like, it is still happy with a less complex sustain phase.
So what the ROLAND MT-32 does is to keep tiny digital snippets of almost 100 instrument sounds in a ROM to create any complexity for the attack phase necessary to reproduce a very realistic sound. After the attack phase, the MT-32 then plays "plain" tunes out of simple digital oscillators for the Sustain phase.
This is nothing new, the Roland D-50 did that, too. But the MT32 can handle 5 independent instrument channels. Also back then the beginning of home computing made digital chips and especially RAM and ROM much more inexpensive so the MT-32 not only appeared to be as good as hundreds of expensive oscillators bundled into a big box (or highly expensive chip), it did even better (5 independent channels) and had only cost a fraction!
So from todays point of view, the MT32 is not very good. You can synthesize hundreds of digital channels with the full tune reproduced instead of just the attack phase with very inexpensive hardware nowadays. But since the MT-32 made the impossible possible back in 1982, it became a legend and so it is still traded with its weight in gold.
You can tell the limits of the MT-32 easily. E.g. a real piano sound requires a complex pattern over the entire length of the tune, the reason why the standard piano sound is very unrealistic played on the MT32. There are other instruments where the attack phase is just not enough to make it sound real. But with almost 100 instruments to choose from (and most very realistic), who cared back then? It was cheap and apparently you got much more than a Pro had for a fraction of the money.